RetouchPi ID Photo: Six Named Sizes Exported at 300 dpi

Why these six: because these are the six people ask for by name. Anything else goes through "Other size…", where the two numbers you type become the specification. Typing 0 mm is refused rather than clamped to 5 mm - a zero-millimetre photo is a mistake, not a smaller photograph.

One full run on 2026-10-09: open a picture, File ▸ ID photo, size "One inch", print "One photo", export. The file that landed was subject-2_25x35mm.png, measuring 295 x 413 pixels - digit for digit the same as 25 x 35 mm at 300 dpi.

All six sizes share one composition: headroom at 12% of the frame height, the eye line at 45%, the shoulders at 85%. The three guides are drawn inside the frame, so when you drag it you are dragging against the specification rather than by feel.

• The frame drags on the photo and scales from its corners: the panel is docked and non-modal, so the pointer is never swallowed by a dialog.
• If you would rather not drag, there are sliders: how much of the picture to take, and moving the frame up/down and left/right.
• "Auto" hands the framing back to what the face detector computed, so a slip is always reversible.
• The guides are a borrowed overlay: open the panel, close it, and no extra lines are left on your canvas.

Framing on a face needs a detector model, and it is loaded on demand (about 233 KB), so the first press of "Auto" shows "Finding a face…". When it finds one it reports: "Face found: the head sits at 12% and the eyes at 45%."

When it cannot, it says it cannot: "No face found, so this export is framed on the selection." The frame then follows your selection and the sliders rather than pretending it aligned a face that is not there. The panel capture above is exactly that state - it is not hidden on this page, because readers will meet it sooner or later, and a page that leaves them puzzled is worse than one they never find.

The backdrop row offers six choices: as it is, white, blue, light blue, red, grey. The panel states the rule itself: "the background is replaced on export only; the pixels in the document do not move."

• Replacing a backdrop also needs a person: when one is recognised it says so ("The subject is recognised: the background is replaced on export"), and when none is it says that instead ("No person recognised, so this export keeps its own background").
• A failed replacement is not silent either: "The background could not be replaced, so this export keeps its own."
• Because the change happens in the export and not in the document, picking the wrong swatch costs nothing - pick again; the picture was never altered.

The print row decides whether the export is one photo or a sheet: a single one, or a layout on paper. Both sheets are whole pixels at 300 dpi: the 5-inch is 1500 x 1050 (5 x 3.5 in), the 6-inch 1800 x 1200 (6 x 4 in).

• When it fits, it counts: "This sheet fits N of them", and the result preview draws that sheet.
• When it does not, it does not squeeze: "This sheet cannot fit the photo, so pick another sheet."
• The preview is honest about one more thing: if a sheet only fits by scaling up, it says the result is upscaled.

• It does not modify the document. An export is an export: the size, the sheet and the backdrop all happen at the moment the file is written, and what is on the canvas is still what you opened.
• It does not guarantee compliance with any one country's rules. What it gives is the size and the composition - millimetres, 300 dpi, headroom 12% / eyes 45% / shoulders 85% - not a ruling on a particular passport regulation. Check the requirements the accepting authority publishes before you submit.
• This page covers layout and export only. Skin and face retouching belongs to a different set of commands, and nothing on this page makes a promise on their behalf.
• Exports are sRGB at 8 bits per channel. A print shop and the web both take that; a formal four-colour print job is not what it is for.